Chemical reactivity in solution emerges from the interplay of molecular structure, reaction partners and environment. Yet, these dependencies are commonly studied one reaction series at a time. This project will investigate whether systematic kinetic measurements across organic reactions can reveal transferable structure in reactivity and provide a quantitative basis for mechanistic hypotheses of chemical behavior. High-throughput kinetic experiments will map condition-dependent reactivity landscapes across electrophiles, nucleophiles and catalytic systems, complemented by molecular descriptors, thermodynamic information and physically motivated kinetic models. Building on these data, machine learning will identify shared reactivity patterns and quantify where they break down, while large language models and agentic systems will be explored to reason over experimental evidence to generate and compare mechanistic hypotheses. Active learning then can select experiments that are maximally informative for resolving uncertain regions of the reactivity landscape or discriminating between competing mechanisms. The resulting measurements can subsequently update both the reactivity model and mechanistic interpretation. The long term goal is a closed loop framework that learns transferable components of reactivity landscapes to predict unmeasured chemical systems.